Rediatric Debate

International Journal of Obesity (2006) 30, 1050–1055. doi:10.1038/sj.ijo.0803331

Variation in physical activity lies with the child, not his environment: evidence for an 'activitystat' in young children (EarlyBird 16)

T J Wilkin1, K M Mallam1, B S Metcalf1, A N Jeffery1 and L D Voss1

1Department of Endocrinology and Metabolism, Peninsula Medical School, Plymouth, Devon, UK





There is currently wide interest in the physical activity of children, but little understanding of its control. Here, we use accelerometers to test the hypothesis that habitual activity in young children is centrally, rather than environmentally, regulated. By central regulation we mean a classic biological feedback loop, with a set-point individual to the child, which controls his/her activity independently of external factors.



Non-intervention, observational and population-based, set in the home and at school.



Girls were systematically less active than boys, and both weekday/weekend day and year-on-year activities were correlated (r=0.43–0.56). A fivefold variation in timetabled PE explained less than 1% of the total variation in physical activity. The activity cost of transport to school was only 2% of total activity, but over 90% of it was recovered elsewhere in the day. The weekly activity recorded by children in Plymouth was the same (to within <0.3%) as that recorded independently in Glasgow, 800 km away. Total daily activity was unrelated to time reportedly spent watching TV.



The correlations within groups and the similarities between them suggest that physical activity in children is under central biological regulation. There are implications both for public health planners and for the potentially novel signalling pathways involved.


physical activity, EarlyBird Study, children, regulation, central control



Energy regulation exacts a fine balance between intake and expenditure. Energy intake is regulated by a complex neuro-endocrine network, centred on the hypothalamus and sometimes referred to as the 'appestat'.1 The effectiveness of appetite control is best appreciated when it fails, as in rare genetic disorders such as Praader–Willi syndrome and leptin deficiency—both characterised by voracious appetite and extreme obesity.2, 3 Energy expenditure combines basal metabolic rate , thermogenic response to feeding and physical activity. Basal metabolic rate and thermogenic response are tightly controlled, but little is known of how physical activity varies in young children, still less of how it is controlled. Given that the opportunity for physical activity varies widely from day to day, we sought evidence from new and published data for an 'activitystat' in young children, to complement the 'appestat' in the control of body weight. An 'activitystat', as we conceive it, would comprise a neuro-humoral feedback loop, with a set-point possibly located in the hypothalamus, able to integrate activity carried out by as yet unknown means, and to control further activity accordingly. If centrally controlled in this way, we would expect overall physical activity to be independent of environmental opportunity or (within limits) of environmental intervention. We therefore tested the null hypothesis that physical activity in young children would not vary according to time, place or opportunity. Evidence for an 'activitystat' and the signalling systems it implies could help our understanding of energy regulation and the current increase in childhood obesity.



We report physical activity in three groups of pre-pubertal school children, one of them carefully randomised (EarlyBird Study), the second selected for range of curricular PE (Three Schools Study) and the third for its different location and culture. Ethical approval was granted for EarlyBird by the S&SW Devon LERC in 1999, and for the Three Schools Study in 2001.

Group 1 consisted of 307 healthy school entrants (137 girls, 170 boys, mean age 4.9 s.d.plusminus0.3 years), drawn from 53 primary schools in the city of Plymouth. They represent the EarlyBird cohort that has been characterised in detail elsewhere,4 and was tested at mean age 4.9 years, and again at 5.9 years. Activity data were complete in 89% and 86% at each age, respectively. Group 2 consisted of a separate set of 215 rather older children (95 girls and 120 boys, mean age 9.0 years (range 6.8–10.6 years)) selected from three schools (S) allocating widely different structured opportunity for physical education in the curriculum (S1: 9.0 h/week; S2: 2.2 h/week; S3: 1.8 h/week) and representing the extremes of socio-economic privilege.5 S1 was a private preparatory school with extensive sporting facilities, S2 a village school with 'Activemark Gold' status for its focus on promoting physical activity in and after school and S3 an inner city school with no special provision or facility. Analysable data were available from 159 (74%) of the children (85 boys, 74 girls). There were no differences in either Group 1 or 2 among those children whose accelerometer data were analysable and those whose were not. Group 3 comprised a randomly selected cohort of 72 healthy children (37 boys and 35 girls, mean age 5.8 s.d.plusminus0.6 years) from the city of Glasgow.6 Glasgow and Plymouth are cities of different size, culture and climate. Glasgow (Scotland; 1.3m population), lies 800 km to the north of Plymouth (England; 0.24m), experiences a different light/dark cycle and significantly colder temperatures, operates a different educational system and traditionally enjoys a different diet.


Materials and methods

Uniaxial accelerometers (Manufacturing Technology Inc., Fort Walton Beach, FL, USA) were used to test the effect of variables that might influence physical activity in children – gender, age, weekday or weekend day, timetabled physical education, location and socio-economic status (SES). The MTI (formerly CSA) accelerometer is widely used and highly reliable.7, 8 It samples movement 600 times a minute and records clock time, duration and intensity of activity. Recordings were made during waking hours over a continuous 7-day period. Activity data were considered analysable only if the accelerometer was worn for a minimum of 4 weekdays and 1 weekend day. Activity counts were analysed in epochs of 1 min, and each minute of the child's waking time designated 'high', 'medium' or 'low' intensity according to the number of counts it contained (low: 0–999 c.p.m.; medium 1000–2499 c.p.m.; high: greater than or equal to2500 c.p.m.).9 It was thus possible to distinguish the time spent in activity of different intensities from the counts (c.p.m.) recorded at each intensity, and relate the two. The data were downloaded onto a PC and areas under the activity curve used to determine total daily activity, weekday activity, weekend day activity, in-school activity and out-of-school activity.

We looked for indications that children maintained consistency in activity over time and distance, that gender differences were systematic in different cohorts and, most importantly, that major environmental variation made no difference to overall activity.

Statistical analysis

All data analysis was carried out using SpSS ver.11.51. Activity was recorded for 7 days (five school days and two weekend days). Each (7-day) sample of physical activity was adjusted for seasonality (number of relevant daylight hours for the week monitored) and for variation between accelerometers (established by a motorised turntable8) using the two respective slope estimates obtained from a multiple linear regression analysis, controlled for age and measures of body fat. Data were analysed only where the child completed at least 4 weekdays and 1 weekend day of recordings. We justified the analysis of children from S1, S2 and S3 (Group 2) together on the grounds that their mean total activities were the same and their variances similar, even when broken down by gender (s.d.=7.5, 6.9, 7.8, 7.4, 7.0, 4.9). The s.d. of 4.9 related to the girls in S1, where numbers were fewest, and was not significantly different from the other variances. The non-parametric Mann–Whitney U-test was performed to confirm, or otherwise, the significance values given by ANCOVA, when subgroups numbered less than 30.



Gender- and age-related differences

If physical activity was centrally rather than environmentally controlled, we reasoned that any gender- and age-related differences should be consistent, irrespective of environment. Accordingly, we compared total weekly activity of Groups 1 and 2, boys and girls separately, and that of Group 1 on two occasions 12 months apart. The activities recorded and gender differences in Group 2 were not significantly different (P>0.05) from those in Group 1 at 4.9 years (Table 1). The girls in Group 1 recorded systematically less activity than the boys, as did the girls in Group 2, and proportionately even less high-intensity activity at both time points 12 months apart.

Variation over time

If control is central, average weekday activity should not differ from weekend day activity despite being differently structured, weekday should correlate with weekend day and average daily activity 1 year should correlate with the next. The mean level of activity in both Groups 1 and 2 was again lower in the girls than the boys at weekends, but unchanged in both genders from weekday to weekend day (Table 2). More importantly, there were substantial correlations (indicating consistency among individuals) between weekday and weekend day activity in both Groups 1 and 2 (Group 1 4.9 years: T r=0.52, G r=0.52, B r=0.51, all P<0.001; 5.9 years: T r=0.56, G r=0.49, B r=0.59, all P<0.001; Group 2: T r=0.46, G r=0.45, B r=0.43, all P<0.001). Total activity did not change from year to year in Group 1 children (37.5 vs 37.4 units, P=0.88), nor the corresponding proportions of high-intensity activity that made up the total (G 29.2 vs 31.2%, P=0.08; B 32.9 vs 33.6%, P=0.56). The correlations for daily activity year-on-year were almost as strong as from weekday to weekend day (T r=0.49, G r=0.36, B r=0.55, all P<0.001).

Environment and physical activity

In order to assess the impact of environmental opportunity on physical activity, we compared the physical activity recorded in children from Group 2 attending three schools that timetabled widely different amounts of PE. As expected, in-school activity was greater in S1 than S2 or S3, in keeping with the greater allocation (Figure 1). However, out-of-school activity was higher in S2 and S3 than in S1 such that, when in-school and out-of-school activity were added together, total daily activity was the same at all three schools (to within 3%). Furthermore, the patterns for total activity (shown) and for high-intensity activity alone (data not shown) were the same for girls and boys. Importantly, less than 1% of the variance in overall activity could be explained by the school environment once the activity data were adjusted for seasonality and the small differences in accelerometer performance identified during QA studies.7 The children at S1 spent 15 h longer each week in school than children at S2 or S3, but only 70 min of it (12 min/day) represented classroom or homework time.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact or the author

Mean in-school, out-of-school and total activity recorded over 7 days by children of mean age 9.0 years (Group 2) in three different schools. S1 provided 9.0 h of timetabled physical education, S2 2.2 h and S3 1.8 h.

Full figure and legend (37K)

Transport to school

Two-thirds (n=185) of the children in Group 1 walked to school, and one-third (n=90) were conveyed by car. This is the national UK pattern, and remains unchanged throughout primary schooling. The non-walkers recorded 16.1% less activity during the hour before and the hour after school, compared with the walkers (Table 3). Although it represented just 2% (0.75/37.5 times 105 c.p.m.) of the activity recorded in a week, 91% of the activity cost of the school run (+0.68 times 105 counts) was 'recovered' in compensatory activity by the non-walkers elsewhere in the day. As a result, there was no difference in total weekly physical activity (0.04 times 105 counts, P=0.97) between those who were driven and those who walked.


In order to test the hypothesis of central regulation in different settings, we compared the activity of EarlyBird children at 5.9 years with published data from a group of similar age from Glasgow, using the same MTI accelerometers (Table 4). Glasgow children recorded the same mean counts (to within <3%) and the same gender differences as the EarlyBird children.

TV/video games

TV watching is popularly associated with low levels of activity, although questionnaire-based data are not unanimous.10, 11, 12, 13 If centrally regulated, total physical activity should not be affected by periods of inactivity. In practice, there was no correlation among Group 1 children between time reportedly spent watching TV/video over 7 days and total activity recorded over the same period (B: r=0.02, P=0.79; G: r=0.04, P=0.72), nor between time spent watching TV/video and low-intensity activity counts (B: r=-0.05, P=0.57; G: r=-0.04, P=0.74). Television in young children did not appear to impact on their physical activity.

Impact of waking time

We investigated the relationships in Group 1 children between waking time and either minutes spent in low-intensity activity or total activity counts recorded, as correlation with the first and lack of it with the second would suggest central regulation to a set-point irrespective of the time available. There were strong correlations between waking time and time spent at low physical activity after controlling for age, seasonality and between monitor differences (G r=0.75; B r=0.74. P<0.001), but none between waking time and total activity counts (B r=-0.04, P=0.69; G r=0.13, P=0.236) or high intensity counts (B r=-0.14, P=0.11;G r=0.04, P=0.69). Thus, those who were awake for longer recorded correspondingly more time in low intensity activity, but a longer day did not add to the amount of high-intensity activity nor impact perceptibly on the total.



Together, the data reported here suggest that children of primary school age display consistency in the amount of physical activity they undertake, independently of opportunity, daily routine, background or culture. Such consistency raises the question of central control.

Control loops are designed to constrain their output within a predetermined range, irrespective of the environment,14 and failure to do so results in the symptoms and signs associated with disease. Biological ranges are determined by two factors(1) differences in loop setting (organised variation) and (2) differences in the quality of loop control needed to meet it (random variation). Organised variation of the set-point admits little environmental effect and should be systematic – that is, reproducible between groups and correlated within groups. Accordingly, the concept of an activitystat as the regulator of physical activity would predict (1) little effect of structured opportunity on total physical activity, (2) systematic variation between groups (e.g. genders, age groups) and (3) correlation within groups (e.g. from time-point to time-point). Further, if the system and its setting are fundamental to the health of the species, the range of the variable it controls might be expected to alter little across locations and cultures. Many, if not all, of these criteria for central control appear to be met in the physical activity of young children.

Although we accept that a comparison of just two cities, even if 800 km apart, is not a rigorous test for difference in location, the striking similarity (within 0.3% overall) in mean physical activity of 6-year-old children between them cannot be readily explained by chance (P=0.92). Self-report risks bias, where accelerometers arguably do not. Our data using accelerometers, suggest that young children from the lower socio-economic groups lose nothing in physical activity through a lack of provision or school-time opportunity. The provision of seven more hours school-time PE per week at S1 made no difference to their overall activity. High-intensity activity varies most among children of this age, and appears more important than low intensity to metabolic health.15 When analysed separately, high-intensity activity was also the same among children attending the three schools. What children lack in school, they make up for – with considerable precision – after school. The systematically lower levels of activity among girls in this and all the other data sets tested, lends support to an activitystat whose setting is lower in girls.

Children from S2 and S3, Group 1 (EarlyBird) and Group 3 (Glasgow) all recorded much the same total activity as those from S1, which timetabled nearly 2 h of physical education every afternoon of the school week. It is worth questioning, how much more can be expected from contemporary children and, by extension, how accurate the popular impression of their inactivity really is. It also seems legitimate to question recently expressed concerns that cutbacks in timetabled physical education in UK primary schools would most affect those of lowest SES,16 as neither we nor others find evidence of social influence on childhood activity.17 The activity that children undertake appears to transcend social status, opportunity and geographical location.

Animal studies are consistent with the 'activitystat' hypothesis. Wheel running in mice is a voluntary activity that can be accurately monitored for duration, intensity, distance and, using a metabolic cage, energy expenditure. Mice have been inbred that record predictably different levels of activity, suggesting genetically different set-points.18 When, by means of an electric drive in a metabolic cage, habitually inactive mice are forced to do more wheel running, they subsequently rest and even reduce their metabolic rate to compensate during the rest of the day. Human twin studies using 'gold-standard' measures have recently estimated the genetic and environmental contributions to physical activity in children. Genetic influence explained 78% of physical activity and 72% of activity-associated energy expenditure in daily life.19

Our study has limitations. Notwithstanding the Glasgow data cited here for comparison, it relates to one geographical area only and needs to be repeated in other settings to be deemed generalisable. It also applies expressly to primary school and not to secondary school children, whose activity patterns may be different. The EarlyBird study, being truly longitudinal, is nevertheless in a position to resolve this question as its cohort moves into secondary school. Although CSA accelerometers have been widely validated for their technical reliability,5, 20 they record uni-axial movement only and are therefore limited in their detection of static work or any extra energy expended in movement against external forces. The similar mean levels of activity from weekday to weekend day, from year to year and between three schools offering very different timetabled physical education, might at first sight be interpreted to mean random data capture. The correlations, on the other hand, suggest meaningful consistency, and the similarities in pattern between the genders that they are robust. The accelerometers were only worn for 1 week, and at different times of the year. However, the collective data from Group 1 represented a whole 12 month cycle, and the data points used in the correlations were strictly 12 months apart. The data on TV viewing is parental report – no worse for that than other data of its kind – but inferior to the objective measure of physical activity. Finally, the numbers studied were small by epidemiological standards. Small numbers might be an issue if we were trying to interpret lack of correlation, but in this study clear correlations emerged to suggest that the activity of children is indeed systematic.

This report tests the hypothesis that variation in physical activity among young children lies with the child, rather than with his/her environment. We propose the term 'activity phenotype' (though it may be genotype) to describe the differences, and can envisage two important implications for the concept. First, the wide range of habitual activity among children may reflect the different set-points of an 'activitystat', a term first coined by Rowland to describe a body sensor that registers and regulates physical work done.21 An understanding of what influences the set-point, and of the neuro-endocrine signals involved, could have implications for managing energy balance as important as those of the 'appestat', where medications capable of influencing it (e.g. sibutramine) are already used in obesity management. Second, simply improving the opportunity for physical activity (engineering the environment) may have little impact on the child whose phenotype is that of the habitually inactive. Like 'horses brought to water', some children – those of intrinsically low activity setting – may simply not participate.

There is little objective data on structured physical activity interventions in children, yet concern has been expressed in the UK for deteriorating sports facilities,22 and the increasing precedence taken by academic over physical education in primary schools. Physical activity has an independent, albeit limited, impact on the metabolic health of children,23, 24 but the means of extending that benefit to more children may be linked, not so much to widening facilities, as to understanding better why some children are satiated at lower levels of activity than others.



Contributions: TJW conceived the EarlyBird study, the Three Schools Study and the Activitystat Hypothesis. KMM undertook the Three Schools Study. BSM was responsible for statistical analysis and ANJ for much of the data collection from the EarlyBird study. LDV is supervisor of the EarlyBird study.



  1. Joliffe N. The appestat; the appetite-regulating mechanism. Merck Rep 1952; 61: 3–7.
  2. Goldstone AP. Prader-Willi syndrome: advances in genetics, pathophysiology and treatment. Trends Endocrinol Metab 2004; 15: 12–20. | Article | PubMed | ISI | ChemPort |
  3. Montague CT, Farooqi IS, Whitehead JP, Soos MA, Rau H, Wareham NJ et al. Congenital leptin deficiency is associated with severe early-onset obesity in humans. Nature 1997; 387: 903–908. | Article | PubMed | ISI | ChemPort |
  4. Voss LD, Kirkby J, Metcalf BS, Jeffery AN, O'Riordan C, Murphy MJ et al. Preventable factors in childhood that lead to insulin resistance, diabetes mellitus and the metabolic syndrome (EarlyBird 1). J Pediatr Endocrinol Metab 2003; 16: 1211–1224. | PubMed |
  5. Mallam KM, Metcalf BS, Kirkby J, Voss LD, Wilkin TJ. Contribution of timetabled physical education to total physical activity in primary school children: cross sectional study. BMJ 2003; 327: 592–593. | Article | PubMed |
  6. Reilly JJ, Jackson DM, Montgomery C, Kelly LA, Slater C, Grant S et al. Total energy expenditure and physical activity in young Scottish children: mixed longitudinal study. Lancet 2004; 363: 211–212. | Article | PubMed | ISI | ChemPort |
  7. Westerterp KR. Physical activity assessment with accelerometers. Int J Obes Relat Metab Disord 1999; 23 (Suppl 3): S45–S49. | Article | PubMed |
  8. Metcalf BS, Curnow JS, Evans C, Voss LD, Wilkin TJ. Technical reliability of the CSA activity monitor: The EarlyBird Study. Med Sci Sports Exerc 2002; 34: 1533–1537. | Article | PubMed |
  9. Metcalf BS, Voss LD, Wilkin TJ. Accelerometers identify inactive and potentially obese children (EarlyBird 3). Arch Dis Child 2002; 87: 166–167. | Article | PubMed | ChemPort |
  10. Hancox RJ, Milne BJ, Poulton R. Association between child and adolescent television viewing and adult health: a longitudinal birth cohort study. Lancet 2004; 364: 257–262. | Article | PubMed | ISI |
  11. Dietz WH, Gortmaker SL. TV or not TV: fat is the question. Pediatrics 1993; 91: 499–501. | PubMed | ChemPort |
  12. Katzmarzyk PT, Malina RM, Song TM, Bouchard C. Television viewing, physical activity, and health-related fitness of youth in the Quebec Family Study. J Adolesc Health 1998; 23: 318–325. | Article | PubMed | ChemPort |
  13. Robinson TN, Hammer LD, Killen JD, Kraemer HC, Wilson DM, Hayward C et al. Does television viewing increase obesity and reduce physical activity? Cross-sectional and longitudinal analyses in adolescent girls. Pediatrics 1993; 91: 273–280. | PubMed | ChemPort |
  14. Wilkin TJ. Endocrine feedback control in health and disease. In: Bittar EE, Bittar N (eds). The Principles of Medical Biology. JAI Press Inc: Greenwich, CT, 1998. pp 1–28.
  15. Sleap M, Tolfrey K. Do 9- to 12 year-old children meet existing physical activity recommendations for health? Med Sci Sports Exerc 2001; 33: 591–596. | PubMed | ChemPort |
  16. Rowe N, Champion R. Young People and Sport National Survey 1999. Sport England:London, 2000.
  17. Kelly LA, Reilly JJ, Fisher A, Montgomery C, Williamson A, McColl JH et al. Effect of socioeconomic status on objectively measured physical activity. Arch Dis Child 2006; 91: 35–38. | PubMed | ChemPort |
  18. Lightfoot JT, Turner MJ, Daves M, Vordermark A, Kleeberger SR. Genetic influence on daily wheel running activity level. Physiol Genomics 2004; 19: 270–276. | PubMed | ChemPort |
  19. Joosen AM, Gielen M, Vlietinck R, Westerterp KR. Genetic analysis of physical activity in twins. Am J Clin Nutr 2005; 82: 1253–1259. | PubMed | ISI | ChemPort |
  20. Trost SG, Ward DS, Moorehead SM, Watson PD, Riner W, Burke JR. Validity of the computer science and applications (CSA) activity monitor in children. Med Sci Sports Exerc 1998; 30: 629–633. | PubMed | ISI | ChemPort |
  21. Rowland TW. The biological basis of physical activity. Med Sci Sports Exerc 1998; 30: 392–399. | PubMed | ISI | ChemPort |
  23. Schmitz KH, Jacobs Jr DR, Hong CP, Steinberger J, Moran A, Sinaiko AR. Association of physical activity with insulin sensitivity in children. Int J Obes Relat Metab Disord 2002; 26: 1310–1316. | Article | PubMed | ChemPort |
  24. Brage S, Wedderkopp N, Ekelund U, Franks PW, Wareham NJ, Andersen LB et al. Objectively measured physical activity correlates with indices of insulin resistance in Danish children. The European Youth Heart Study (EYHS). Int J Obes Relat Metab Disord 2004; 28: 1503–1508. | Article | PubMed | ChemPort |


Grants: The study was supported by grants from Diabetes UK, S&SW NHS Executive R&D, The Diabetes Foundation and the Child Growth Foundation. We gratefully acknowledge further help from Roche Products, Smith's Charity, Abbott Laboratories, Ipsen, GSK, Astra-Zeneca, Unilever and the Beatrice-Laing Foundation.



These links to content published by NPG are automatically generated

Extra navigation